|
|
 |
| Scalar model outperforms configural m... |
 
|
| Message/Author |
|
|
|
|
Dear Mplus community.
I am doing an ESEM validation of an psychometric instrument.
I want to check for measurement invariance. I have a clustering variable and my items are categorical. Therefore I use want to test a unconstrained model against a scalar model.
However when I try to do so, the scalar model achieve better fit than the unconstrained (configural) model. How can this be?
Results are as follows:
agegroup gender level
chi-squaretest 467.747*** 266.665*** 469.153***
CFI config 0.961 0.968 0.957
CFI scalar 0.971 0.975 0.963
delta CFI -0.010 -0.007 -0.006
TLI config 0.956 0.963 0.951
TLI scalar 0.973 0.976 0.966
deltaTLI -0.017 -0.013 -0.015
RMSEA config 0.043 0.04 0.04
RMSEA scalar 0.033 0.032 0.033
delta RMSEA 0.010 0.008 0.007 |
|
|
|
|
| If you are using WLSMV, only the special difference testing can tell you how the models compare, not the chi-squares themselves. If this doesn't help, we need to see the full output of these runs to diagnose it - send to Support along with your license number. |
|
|
| Back to top |
|
|
